| Title | Development of prevention strategies for silica scaling based on neural network supervised machine learning |
|---|---|
| Authors | S. Juhri, K. Yonezu, R. Terashi, T. Naritomi, E. Watanabe, S. Sato, N. Inoue, T. Yokoyama |
| Year | 2025 |
| Conference | New Zealand Geothermal Workshop |
| Keywords | machine learning, silica scale, mitigation, inhibitor |
| Abstract | Scaling is a persistent challenge in geothermal energy extraction, affecting the subsurface and surface facilities of geothermal power plants. However, a universal prevention method has not been established due to the complexity of scale formation and diverse types of scale deposition. Recently, artificial intelligence (AI) has been vastly used in various disciplines of science, including geothermal exploration and optimization of energy utilization. Our previous work attempted to utilize supervised machine learning (SML) to accurately predict the formation rate of silica scale in a geothermal environment. The produced models yielded error values of < 0.15, as root mean squared error (RMSE), suggesting the applicability of the models to predict silica scale formation. In this study, we utilize the SML models to develop mitigation strategies against silica scaling. The strategies were developed based on the contribution factors of the chemical parameters of geothermal water. Then, three scenarios were evaluated: (1) pH modification only, (2) pH modification with removal of metals, and (3) pH modification and polymerization of oversaturated silicic acid. In addition to the modeling, onsite experiments of batch metal immersion were also conducted at the Hatchobaru geothermal power plants to examine the strategy and the development of a new silica scale inhibitor. The experiments were conducted at a temperature of ~90 °C, simulating the aging tank's condition. This study exhibits the potential of a machine-learning model to predict and assist in the mitigation of silica scale formation. With further and more experiments in more diverse properties of geothermal water, future models are expected to be applicable to more geothermal fields. |